Knitting has accompanied humanity for millennia: from imperial robes to grandmother's sweaters, from handcrafted hats to lovingly made scarves. It is an ancient knowledge, passed down from generation to generation, made of patience, mastery and creativity. Or at least, it was.
A team of Laurentian University in Canada has just announced an artificial intelligence system capable of โlookingโ at a fabric and automatically replicating it. The expert eye of the craftsman is no longer needed, nor his manual skills: a smartphone and a robot are enough. Automation has already devoured many traditional professions, but this time it seems to have targeted one of the most intimate symbols of human creativity, with potentially devastating consequences for millions of workers.
The magic of the algorithm that โseesโ the points
Remember when turning a picture into a knitted piece of work required hours of patience? Someone had to carefully examine the photo, identify each stitch, and translate it into instructions that a machine could understand. It was a slow, laborious, and hopelessly human process.
Canadian researchers Xingyu Zheng e Mengcheng Lau they have turned this scenario upside down. Their deep learning model works in two stages: first simplifies the image of the fabric, transforming it into a sort of diagram that highlights the visible points; then deduces the complete instructions for the machine, including hidden layers. It is as if the algorithm has learned not only to see, but to think like a master knitter.
The results speak for themselves: tested on approximately 5000 textile samples, the model achieved an accuracy of over 97%. Even multi-colored threads and the rarest stitches, which were insurmountable obstacles for previous methods, are interpreted correctly. Manual translation now seems to be a thing of the past.
Between Cultural Preservation and Mass Unemployment
Innovation always has two sides, and this is no exception. On one side, the possibility of preserving and reproducing ancient textile designs has an inestimable cultural value. Let us think of the traditional patterns that are at risk of disappearing, of the artisanal techniques that could be documented and replicated with unprecedented precision.
On the other hand, however, the shadow of unemployment looms. The textile industry employs over 75 million people worldwide, many of whom are engaged in tasks such as knitting and sewing. In countries such as Bangladesh, Myanmar ed Ethiopia, entire communities depend on these jobs for survival.
Our model achieved over 97% accuracy in converting images into knitting machine instructions, significantly outperforming existing methods.
I wanted to quote verbatim the (cold) statement of the co-authors of the study, Haoliang Sheng e Songpu Cai. Behind this percentage lies a potential social tsunami: if machines were to take over on a large scale, millions of families could lose their livelihoods, worsening economic hardship in regions already marked by poverty and limited opportunities.
Knitting: The Race Towards an Automated Future
The researchers are already planning to further improve their deep learning model, aiming to test it on more complex sewing tasks. The study, published in the magazine Electronics, represents only the beginning of a textile revolution.
I wonder, though, if there is room for broader considerations in the journey to this technological marvel. Algorithms learn to replicate human mastery, but they cannot replicate the social value of work. They can recognize dots and patterns, but they do not see the lives that depend on those repetitive gestures.
The promise of a fully automated textile production, which reduces time and costs, hides a fundamental question: at what price? Technology advances inexorably, while the time to find socially sustainable responses it gets shorter.
Just like a piece of knitting that, once undone, cannot return to the original ball.
